The All-New Switch Book: The Complete Guide to LAN Switching Technology, Second Edition

13.4. Getting into the Priority Business

Let's explore a fundamental question regarding the use of priority in switched catenets:

At what point do you classify and differentiate traffic based on priority?

There are two parts to the problem:

1. Assigning a priority to a given frame or stream of frames.

2. Providing preferential treatment across the catenet based on the priority assignment.

In general, priority classification is more difficult once you are removed from the application and end station sourcing the traffic. That is, if an application itself is priority-aware, it can indicate its requested user priority in all of the traffic that it originates. There is no need to infer the priority from the contents of the frame. If an application (or the end station on which it is running) is not priority-aware, then priority-enabled switches will have to determine the desired priority by inspecting each frame's contents and applying some set of administrative rules. This operation can impose a considerable performance penalty in switches, especially if such classification must be performed millions of times each second.

The sooner in the end-to-end communication process you make the priority classification, the sooner you can provide preferential treatment, and the easier it becomes to implement high performance priority-enabled switches. Once a priority determination is made, the frame can be tagged with that priority, and all devices down the line do not need to know the policy rules ...